A dataset of land use types in the Lijiang River Basin from 2016 to 2020

Abstract

Land use monitoring is an important basis for land resource management. On the basis of Sentinel-2 remote sensing image data, we used multiple preprocessing methods such as original data calibration, Mosaic and tailoring, and established remote sensing interpretation markers as classification training samples. Then we adopted the maximum likelihood method to classify the land use situation of the Lijiang River Basin in Guilin City from 2016 to 2020. Moreover, in combined with the original random test samples, we used the confusion matrix analysis method to evaluate the accuracy of the dataset. The classification results show a relatively good classification effect of land use types in the Lijiang River Basin, with an overall classification accuracy above 95%, and the kappa coefficient above 0.8. On the whole, the land use types in the Lijiang River basin changed little, and the main land use type is woodland, accounting for about 65.2%. The spatial heterogeneity show that the construction land area expanded from 4.17% to 5.19% from 2016 to 2020. This dataset can provide data support for the study of spatio-temporal changes of land pattern and land use transfer and evolution.